On this episode of Navigating Forward, Launch Consulting's Kevin McCall, Managing Director of AI, and Melanie Roberson, Director of Organizational Effectiveness, talk about the 5 Ds of AI: dull, dirty, dangerous, difficult, and double-time. The 5 Ds framework is a way of thinking about the type of tasks where AI can complement and augment human capabilities at work. With examples of each of the 5 Ds, they delve into how AI that's guided by humans can empower and support us in our roles, so that we have more time to spend on more meaningful tasks. Listen in to hear about how the collaboration between AI and humans plays into the future of work.
Find Kevin at https://www.linkedin.com/in/kevin-mccall-74b66/
Find Melanie at https://www.linkedin.com/in/melanie-roberson-55a0a23/
Learn more about Launch Consulting and AI at https://www.launchconsulting.com/ai-first
00:00:03:09 - 00:00:45:19
Welcome to Navigating Forward, brought to you by Launch Consulting, where we explore the ever-evolving world of technology, data, and the incredible potential for artificial intelligence. Our experts come together with the brightest minds in AI and technology, discovering the stories behind the latest advancements across industries. Our mission: to guide you through the rapidly changing landscape of tech, demystifying complex concepts and showcasing the opportunities that lie ahead. Join us as we uncover what your business needs to do now to prepare for what's coming next. This is Navigating Forward.
00:00:45:22 - 00:00:56:21
Hello, everyone. My name is Kevin McCall, and I am delighted to be here today with Melanie Roberson to discuss the Five Ds of AI. So, Melanie, please tell everyone a little bit about your role here at launch.
00:00:56:29 - 00:01:11:04
Thanks, Kevin. Hi, everyone. My name is Melanie Roberson, and I am the Director of Organizational Effectiveness here at Launch. I come with 20 plus years of experience helping organizations thrive, primarily through people strategy. Kevin, tell us about you.
00:01:11:07 - 00:01:24:05
All right. Yes. I'm the Managing Director of AI here at Launch Consulting, and prior to Launch, I spent almost 26 years at Microsoft, most recently managing a worldwide team that focused on AI for autonomous systems.
00:01:24:07 - 00:01:29:25
So impressive. Tell us more about the Five Ds framework. When were you first introduced to it?
00:01:30:00 - 00:02:14:25
Well, I was first introduced to this concept in an MIT robotics class where they talked about the 3 Ds of robotics, where robots are often used for jobs that are dull, dirty, and dangerous. And the gentleman who facilitated my discussion group also added the fourth, he referred to this as double-time tasks. But over the last several years, as I've been exposed to many different AI projects, I realized that every one of these applies really well to AI, actually, and I had added a fourth category as well. I called that difficult because there are also a whole bunch of tasks that fall more uniquely into the difficult category. So, I felt like it was worthy of inclusion on the list of task types that are really great for AI.
00:02:14:28 - 00:02:17:13
This is such a great list, Kevin. Tell us more.
00:02:17:16 - 00:02:32:26
So, whenever I talk about this list, I frequently stress the task aspect of these jobs and the reason I stress task is because there's a general fear and uncertainty around how AI will replace jobs or eliminate roles. I don't know if you've seen that.
00:02:33:01 - 00:02:34:03
I've definitely seen that.
00:02:34:04 - 00:03:32:13
Yeah, especially lately, of course, with ChatGPT and how since ChatGPT has become really part of the public consciousness, I think people are thinking about this more. And while there are certainly some very narrow circumstances where I think AI may indeed eliminate roles, I think almost without exception, I don't believe AI is going to replace people or roles.
I think it's going to be more focused on more narrow tasks. So, think about it in terms of logical subsets of roles, right? And I think that this work will be performed under the supervision and guidance of humans most of the time anyway. So, every single one of us performs many different types of tasks in our jobs, and for some jobs, they might be only able to help with five or ten percent of those tasks. But in others it might be 30, 40, maybe even half of their jobs. But in my experience, it's really, really rare that AI is going to be able to perform all of the tasks of a given role.
00:03:32:13 - 00:03:46:07
Yeah, that makes sense, Kevin. So, AI is really changing how work gets done and who does it, but it's really focused on complementing and augmenting our capabilities. It's not really replacing them, as some people might think.
00:03:46:08 - 00:04:03:19
Yeah, absolutely. As a matter of fact, I think an argument could be made that that even though there might be very, very narrow roles that could be eliminated by AI, I think an argument could be made that the net result will be more jobs created than lost as part of this AI wave that we're seeing.
00:04:03:22 - 00:04:44:06
Yeah, you know what? This is really in line with my thinking because it reminds me of a recent study that was done by an AI team and some of the general findings in that study were that the organizations that have had the most significant performance improvement leveraging AI were the ones that were really focused on humans and machines working together. And AI really complementing the task and not supplanting them. And so, this idea that humans and AI can actively work together to enhance each other's strengths is really the idea that is scalable across organizations and across companies as thought of in this study.
00:04:44:10 - 00:04:48:03
Yeah, that's well said. I think I saw that MIT study, so I think that's a good one.
00:04:48:08 - 00:05:08:10
Yeah. I think the important idea here is that companies must really understand how humans can most effectively augment machines and how machines can enhance what humans do best and how, you know, both that symbiotic relationship can really be seen as a partnership and not kind of an opportunity for elimination.
00:05:08:15 - 00:05:39:02
That makes sense. And that's actually very consistent with what I saw at Microsoft before I joined Launch. I probably had 600 or so customer meetings with over 100 unique companies in the last five years I was there, and I only encountered one company who explicitly said that we’re going to need to see headcount reduction as a precondition for funding this project. So generally speaking, I never saw that. I never saw organizations focusing, even in autonomous AI, I never saw organizations saying, we've got to eliminate headcount.
00:05:39:07 - 00:05:48:03
That's great. I mean, even with that kind of conventional wisdom out there, that's really that's really great to see. So, tell us about the “dull” on your list.
00:05:48:05 - 00:06:26:15
Okay, we’ll start with full. So dull is probably the simplest and frankly, most obvious task type that applies in this case, of course, most frequently to office workers. So, think of all those very tedious, repetitive, administrative tasks that are present in every organization. You know, rebate and claims processing, billing payments, invoicing, new account setup. Obviously, we could go on and on. This applies to all industries. And while classic automation approaches, or what many people have recently started calling RPA projects, apply here, AI technology is being used more and more often to deliver a greater efficiency on these types of tasks.
00:06:26:15 - 00:06:29:00
So first I'll ask, what is RPA?
00:06:29:03 - 00:06:50:01
Oh yeah. RPA stands for Robotic Process Automation, and it actually does not involve robots. It's a metaphor in this case but think of it as paper processing. All those administrative tasks could have to do with paper, you know, these highly repetitive processing of various types of, you know, of documents that every organization has to perform.
00:06:50:03 - 00:07:22:13
Well, this makes perfect sense. And this is a great example of AI and humans working together because I think one of the core strengths of technology is its ability to complete task at scale and error free. And I think dull tasks leave themselves open to error because, well, they're boring. So, this is a great example of optimizing human capabilities to do kind of better, more meaningful tasks, because why use humans to do dull work when they could be performing work that's really more meaningful and more value added?
00:07:22:13 - 00:07:36:26
Right, right. And as a matter of fact, the Microsoft Future of Work study. I don't know if you saw that, but most workers agree with that, too, where they want their jobs to be more meaningful. They want to hand off more of their stuff to AI so they can spend more time on more meaningful tasks.
00:07:36:27 - 00:07:41:10
They're requiring it, you know, the generation, the future of work generation is requiring more meaningful work.
00:07:41:10 - 00:07:41:28
They expect it.
00:07:42:21 - 00:07:43:24
Yeah, a lot of them expect it.
00:07:43:26 - 00:07:49:02
That's great. So, could you tell us more about the recent experience you've had with this in customer calls?
00:07:49:02 - 00:08:24:07
Yeah, I know I mentioned this to you a few days ago. I'm going to give you one literally from last week. I just moved into a new townhome here in Seattle. It's in Queen Anne, overlooking the city. We really like it. So, I went to the website to activate my natural gas account and the website says this is not a valid address. So, I called them, and I said, you know, here's my address. And the person said, well, you don't have natural gas at that address. And I said, no, I promise you I do. I boiled water this morning. And oddly, we went back and forth on this several times and where we ended up is I went and took a picture of my natural gas meter, and I emailed it to them.
00:08:24:11 - 00:08:26:01
Did you take a picture of your boiling water?
00:08:26:02 - 00:09:24:04
No, I didn't, just the meter, just the meter. They had to believe it was mine. And then they read the serial number of the natural gas meter off the tag that they had actually placed on the meter. And they said, oh right. The developer who built your townhome had listed it under a different address.
So, the problem was solved. But the thing that struck me was that this didn't require a person, right? A person was not necessary to do this. I could have uploaded a picture to their website, and it would have been trivial for a neural network-powered OCR, you know, software package to read the serial number right off of that image and set up the account for me.
So, there are certainly many, many more mainstream examples with wider applicability. But this is fresh in my mind because I had to deal with this last week. And it's a great example of where very simple in this case, you know, a cognitive service in Azure could have easily pulled the serial number right off the natural gas meter and set up the account for me.
00:09:24:08 - 00:09:29:09
Yeah, we have all been on those calls. So, I have to ask, what is OCR?
00:09:29:14 - 00:10:03:23
Oh, yeah. So, OCR in this case, optical character recognition. So, it's a situation where it can process, of course, let's say a PDF to read the text right off the page or in this case you can also use OCR to pull images or pull characters, numbers, text strings right out of an image, right. In this case, it could have pulled the, you know, the actual serial number of the natural gas meter itself, you know, as well as the unique identifier that was put on there by the local utility company.
00:10:03:28 - 00:10:07:16
So, another example of AI just making things easier for humans.
00:10:07:16 - 00:10:08:03
00:10:08:07 - 00:10:11:29
Great. So now let's move on to difficult. What's difficult?
00:10:12:00 - 00:12:11:10
Yeah, so difficult. And I'm pleased to add this one in here, I think it's a very large bucket. I think the best examples here are decisions that humans must make based on consideration of, you know, various input variables. So, to be more specific, there are near infinite number of decisions that knowledge workers must make every day in every industry on a subset of variables that get them close enough to making optimal decisions.
So, you know, stock trading, staffing, industrial control, logistical optimization, the list really is endless. Now, if they aren't using computers, most domain experts establish heuristics for making these decisions, right. And they consider normally a single digit number of variables, you know, four, six, maybe eight, because of either the complexity or the time required to consider more, it's hard to mentally process many more variables than that, and they might not have the time to consider more anyway. So, using computers clearly would make it easier to consider more variables more quickly. But still, someone has to create that Excel spreadsheet or the model in order to consider the interplay between all these variables. So, you still end up with heuristics and software that use maybe five, ten, maybe 15 input variables that allow the users of those applications to approach, you know, optimal decisions.
So, AI is increasingly being used to make more optimal decisions based on far more variables, far more quickly. And sometimes these AI agents learn from data directly, and sometimes they learn from experience by interacting with models that have been created of a given domain. But either way, AI is becoming more prevalent and more autonomous in its ability to control robots, fly drones, control industrial equipment, you know, or even directly control, you know, things like chemical reactors in ways that far surpass what humans can do in terms of complexity or time.
00:12:11:12 - 00:12:22:08
So that's very interesting Kevin. And so, while computers are really great at managing this complexity, would you still say that it's humans that are guiding the managing of this complexity?
00:12:22:08 - 00:12:37:12
Oh, absolutely, yeah. As a matter of fact, I consider that assumed in my experience the involvement of and the deep collaboration with the appropriate subject matter experts is always required in order to deliver the intended business result with AI.
00:12:37:19 - 00:12:40:16
So now let's talk about dirty.
00:12:40:18 - 00:13:09:19
Okay. Sure. Yeah, this is a domain where autonomous devices like robots or drones are well-suited to handle things that are just unpalatable for humans. So, things like processing garbage or recycling, maybe an autonomous watercraft that's cleaning up an oil spill or something like that. Sometimes these systems are mobile and sometimes they're fixed, but there are more and more AI-powered systems that can perform tasks in environments that are that are just unappealing for human workers to work in.
00:13:09:26 - 00:13:20:00
And I really love the idea of just enabling us to do work that we wouldn't be able to do otherwise, really important work that we wouldn't be able to do otherwise.
00:13:20:03 - 00:13:21:14
00:13:21:17 - 00:13:24:00
So now let's talk about dangerous.
00:13:24:02 - 00:14:16:17
Sure. And this is, there's kind of a continuum between this one and the last category. So, this also includes tasks at which robots and drones are very well suited. So, consider like a wide wheel-based robot that goes into a mine shaft where there may obviously be a risk of collapse, but there may also be noxious gases present, right? Or using a robot or a drone for first trips into spaces where the sensor packages allow navigation without visible light. So, you know, these kinds of examples make perfect sense. First responders, another great example where first responders for fires or building collapses or let's say a bomb threat investigation or bomb disposal. You know, there are many examples where we would much, much rather have agents perform these tasks, either on our behalf or under our control, under our guidance.
00:14:16:21 - 00:14:30:07
Right. Yep. Again, important work where AI technology just expands our abilities to be able to perform. So, let's talk about the last one, double-time.
00:14:30:11 - 00:15:23:07
Right. So, there are lots of examples where you either need greater speed, greater precision with high consistency, more consistent movement with heavy payloads, or some combination thereof. So, an obvious example that, you know, is quite familiar to people is something like automotive manufacturing, right? Where robots not only move fast, but they move extremely consistently. So, there's not only a huge productivity benefit, but also higher first-time quality because of the consistency of the movement of the robots.
But there are other repetitive motion situations like, let's say, having a fixed-mounted, six-degree-of-freedom robot unloading and loading heavy bags of coins from carts, let's say in a bank or a government facility. So, having a robot do that allows you to worry a little bit less about things like OSHA violations, right?
00:15:23:13 - 00:15:30:24
That's great. So, in conclusion, Kevin, what are our takeaways from your list of five?
00:15:31:00 - 00:16:32:07
Sure. There are two reasons why I really like this Five D framework. The first is that many companies I talk to are still trying to figure out where to start with AI and thinking about specific tasks that might be good candidates for AI. It's not only easier to think about roles or jobs, but it also allows you to crowdsource the best ideas directly from your workers by just asking them, you know, what are the tasks that you do every day that you'd love to hand off to an AI under your supervision that would allow you to get more done and allow you to enjoy your job a lot more.
And frankly, that's a good segue to the second reason I love this framework is because thinking about AI performing these kinds of tasks reinforces the idea that AI really is best used in empowering, supporting, augmenting people. We obviously believe that strongly here at Launch. But as you know just as well, if not better than I do, Melanie, recent research, as you've mentioned, has also reflected that as well.
00:16:32:12 - 00:16:56:06
Yeah, and I'm glad you brought that up. As I shared before, the effective use of AI will come from this symbiotic relationship between technology and humans. And we already know that AI will radically alter how work gets done and who does it. And we know that the large impact will be really focused on complementing and augmenting human capabilities and not replacing them.
00:16:56:11 - 00:17:13:04
Absolutely. I absolutely agree with that statement, Melanie, and that's probably a perfect place to finish our conversation for today, though, something tells me that a future podcast on the top x list that represents the attributes of a great overall AI program may be in order.
00:17:13:04 - 00:17:19:01
Yeah, I think that's a great idea. All right. Well, join us next time on Navigating Forward.
00:17:19:03 - 00:17:19:12